How are we doing ensembles?

[ljwolf]

# object-oriented
my_model = create_NPZD_model(
    ...
)

my_ensemble = [
    revise_model(my_model, :min_growth, 8 + normal(0, .1))
    for _ in range(1000)
]

respec_* # create a copy with an adjusted set of attributes

respec(my_model, :field, value)

respec_phyto_args(my_model, 
    :field, value
)

my_sNPZD = respec(
    my_model,
    :predation_function,
    my_exotic_random_predation_function
)

# this operates in place
set!(my_model, :field, value)

## Alternativey hard-coded random initial states can be used, too, to set up an ensemble

my_ensemble = [
    create_NPZD_model(
    phyto_args = Dict(
        "min_growth" => 8 + normal(0,.1)
    )
    ...
) for _ in range(100)
]

my_ensemble_job = agate_ensemble(
    my_ensemble,
    pool = 
)

# under the hood, we use Distribute for 
# the computation for the ensemble

To execute this in parallel, we would define some simulate() function that uses the @parallel macro. See the slurm+Distributed example here.